Page History

If a floating-point value is to be demoted to a floating-point value of a smaller range and precision or to an integer type, or if an integer type is to be converted to a floating-point type, it the value must be ensured that this value is representable by represented in the new type.

Section 6.3.1.4 of C99 says \[[ISO/IEC 9899:1999|AA. C References#ISO/IEC 9899-1999]\] says:

When a finite value of real floating type is converted to an integer type other than _Bool, the fractional part is discarded (i.e., the value is truncated toward zero). If the value of the integral part cannot be represented by the integer type, the behavior is undefined.)

When a value of integer type is converted to a real floating type, if the value being converted can be represented exactly in the new type, it is unchanged. If the value being converted is in the range of values that can be represented but cannot be represented exactly, the result is either the nearest higher or nearest lower representable value, chosen in an implementation-defined manner. If the value being converted is outside the range of values that can be represented, the behavior is undefined.

And section 6.3.1.5 says \[[ISO/IEC 9899:1999|AA. C References#ISO/IEC 9899-1999]\] says 

When a double is demoted to float, a long double is demoted to double or float, or a value being represented in greater precision and range than required by its semantic type (see 6.3.1.8) is explicitly converted (including to its own type), if the value being converted can be represented exactly in the new type, it is unchanged. If the value being converted is in the range of values that can be represented but cannot be represented exactly, the result is either the nearest higher or nearest lower representable value, chosen in an implementation-defined manner. If the value being converted is outside the range of values that can be represented, the behavior is undefined.

Consequently, in implementations that do not allow for the representation of all numbers, conversions of too-small numbers ( between zero and FLT_MIN) may result in undefined behavior.

This rule does not apply to demotions of floating-point types on implementations that support signed infinity, such as IEEE 754, as all numbers are representable.

Noncompliant Code Example (int-float)

The following This noncompliant code example leads to undefined behavior if the integral part of f1 cannot be represented as an integer.

float f1;
int i1;

/* initializationsinitialize fl */

i1 = f1; /* Undefined if the integral part of f1 > INT_MAX */

...

This compliant solution assumes that the range of floating-point values is greater than that of integers an int(this is the case in almost all implementations). Unfortunately, there is no safe way to inquire about this assumption in the code short of already knowing the implementation.

float f1;
int i1;

/* initialize initializationsfl */

if (f1 > (float) INT_MAX || f1 < (float) INT_MIN) {
  /* Handle Error */
} else {
  i1 = f1;
}

Noncompliant Code Example (demotions)

The following This noncompliant code shows casts of values example contains conversions that may not be outside of the range of the demoted typetypes.

long double ld;
double d1;
double d2;
float f1;
float f2;

/* initializations */

f1 = (float)d1;
f2 = (float)ld;
d2 = (double)ld;

...

Compliant Solution (demotions)

This compliant solution properly checks to see whether the values to be stored can be represented in the new type.

#include <float.h>

long double ld;
double d1;
double d2;
float f1;
float f2;

/* initializations */

if (d1 > FLT_MAX || d1 < -FLT_MAX) {
  	/* Handle error condition */
} else {
  	f1 = (float)d1;
}
if (ld > FLT_MAX || ld < -FLT_MAX) {
	  /* Handle error condition */
} else {
	  f2 = (float)ld;
}
if (ld > DBL_MAX || ld < -DBL_MAX) {
	  /* Handle error condition */
} else {
	  d2 = (double)ld;
}

Risk Analysis

Failing to check that a floating-point value fits within a demoted type can result in a value too large to be represented by the new type, resulting in undefined behavior.

...